Hi Dan,

I agree with Engin that you should not need to generated a special cif for 
regular N-glycan for REFMAC. In the past, REFMAC gave me the impression that 
it handles N-glycans very smoothly. And the only times that it compained.. 
[skip to the end of this email to see this].


So I did the following test:

I downloaded the IgG1 Fc structure 3SGK from PDB, pdb3sgk.ent and 
3sgk-sf.cif, then coverted the 3sgk-sf.cif to mtz by "cif2mtz hklin 
3sgk-sf.cif hklout 3sgk.mtz".

Test 1): restrained refinement, pdb3sgk.ent against 3sgk.mtz: PASS.

Test 2): remove all header of pdb3sgk.ent till CRYST1 and all CONNECT items 
at the end, generating pdb3sgk_noheader.ent, then
restrained refinement, pdb3sgk_noheader.ent against 3sgk.mtz: FAIL.

Now refmac complains that it found new ligand and generated a small cif file 
"test_17_lib.cif" and suggested me to check it. Interestingly, the new cif 
file describes an ASP-NAG linkage, instead of ASN-NAG. A look at the pdb 
file shows that it is actually linking NAG 1461 and ASP 64 of chain C, two 
residues got too close in the original PDB. (How lucky I was to find this 
good example by chance!)

So, two more tests:
Test 3):
restrained refinement, pdb3sgk_noheader.ent against 3sgk.mtz, with the 
additional test_17_lib.cif : PASS, but at the cost of recognizing a linkage 
that is probably not natural (1.3A in the resulting PDB file, and is 
formally recorded in the resulting pdb file header):
##refined pdb:
LINKR        OD1 ASP C  64                 O6  NAG C1461 
ASP-NAG
####REFMAC log
  WARNING : link:ASP-NAG  is found dist =     1.335 ideal_dist=     1.309
            ch:CC   res:  64  ASP              at:OD1 .->Cj   res:1461  NAG


Test 4): simply rotate O6 of NAG 1461 away from ASP 64 by editing the chi 
angle of C5-C6, to generate pdb3sgk_NAG1461fix.ent
restrained refinement, pdb3sgk_NAG1461fix.ent against 3sgk.mtz: PASS
So, it holds true that REFMAC does not need additional information for 
N-glcyans, as long as all atoms are at correct positions.



Now back to where I said earlier:
....and the only times REFMAC complained.....

In the past, the only times that REFMAC compained to me about N-glycan, is 
when the N-glycan has atoms placed too close (within covalent bonding 
distance) to other atoms that it should not have covalent bond with. In such 
cases REFMAC would guess that there is a novel covalent bond to be made and 
it will try to generate a new cif and ask the human to check its validity. 
In the 3SGK case, the ASP-NAG link for ASP64 - NAG1461 on chain C was caused 
by their O6 and OD2 atoms coming too close. Unfortunately, this linkage was 
recorded in the original pdb3sgk.ent in the header, and that's why test 1 
could PASS but test 2 couldn't.
line 803 LINK         OD1 ASP C  64                 O6  NAG C1461     1555 
1555  1.34


[Conclusions]

a) REFMAC does not need extra description for regular N-glycan structures, 
given that none of the atoms of the N-glycan is placed too close to other 
atoms by mistake.
b) REFMAC makes guesses of covalent bonds to ligands based on atom-atom 
distances. Ligand atoms getting too close to other atoms will trick REFMAC 
to believe that a new linkage or a new compound is encountered, and abort.

HTH

Zhijie













-----Original Message----- 
From: Dan Freed
Sent: Monday, July 27, 2015 3:33 PM
To: [log in to unmask]
Subject: Re: [ccp4bb] N-linked Man3GlcNAc2 cif file

Thanks for your quick replies, Christian and Engin.  I'm not sure if I 
explained my problem very well.

The problem is not with building the sugar chain, it is with refinement in 
Refmac.  I have built the pentasaccharide successfully, and I have also 
specified the proper "LINK" and "LINKR" designations in the pdb file.  Coot 
has no problem recognizing these linkages, and performing real space 
refinement of the pentasaccharide.  However, Refmac cannot handle it and 
fails every time.  The .cif file that Refmac makes by default does not work 
in subsequent runs, either.

Maybe I should just switch to Phenix for the remainder of the refinement 
process...



----- Original Message -----
From: "Engin Özkan" <[log in to unmask]>
To: "Dan Freed" <[log in to unmask]>, [log in to unmask]
Sent: Monday, July 27, 2015 2:33:29 PM
Subject: Re: [ccp4bb] N-linked Man3GlcNAc2 cif file

I am not that familiar with Refmac, but I don't think you should be
defining sugar chains as one cif. Refmac uses the standard monomer
library (as Phenix did), which knows the sugars and the types of links
contained in N-linked sugars (and other ones, too). I checked the Refmac
manual, and the interesting keywords there appear to be LINK and SUGAR.

The list of links known to Refmac are here:

http://www.ccp4.ac.uk/html/refmac5/dictionary/list-of-ligands.html#links

Among those, "NAG-ASN" would define the first Asparagine to N-acetyl
glucosamine. "BETA1-4" would link NAG#1 to NAG#2, "BETA1-4" would link
NAG#2 to BMA, "ALPHA1-3" and "ALPHA1-6" would link the BMA to the MANs, etc.

Engin

On 7/27/15 12:28 PM, Dan Freed wrote:
> Hi all,
>
> I'm refining a protein in Refmac with several N-glycosylation sites. 
> Refmac was able to recognize and successfully write a .lib parameter file 
> for sites that had density for the first three sugars (NAG-NAG-BMA), and I 
> could use this file in Coot and Refmac without any issues.  However, I am 
> having trouble making a .cif or .lib file for glycosylation sites that 
> have density for the next two sugars (1,3- and 1,6-alpha glycosidic 
> MAN-BMA linkages, forming the core pentasaccharide Man(3)GlcNAc(2)).
>
> I've numbered the residues consecutively (no overlap with protein 
> residues), designated all of the sugar atoms as HETATM, and specified LINK 
> lines in the pdb file.  However, Refmac still fails, and won't write a 
> .lib file that I can use in subsequent runs.  I haven't had luck with .cif 
> files generated from PRODRG, can't find a file in HIC-Up or Ligand Expo, 
> and I'm not entirely sure how to run eLBOW on more than one 'ligand', 
> since I technically have a NAG, BMA and MAN oligosaccharide.
>
> I'd be very grateful if someone could explain how to make (or find) a .cif 
> file for N-linked Man(3)GlcNAc(2).  Thanks in advance!
>
> Best,
> Dan


-- 

Daniel M. Freed, PhD
NIH Postdoctoral Fellow
Lemmon lab
Department of Biochemistry & Biophysics
Perelman School of Medicine
University of Pennsylvania
Philadelphia, PA
lab: (215)898-3411
mobile: (910)367-6485
email: [log in to unmask]
www.med.upenn.edu/lemmonlab